Transcriptome Analysis Reveals Neuroprotective Aspects of Human

Home , ACTG1, HADHA, MYL6, MYL6B, POLR2E, PPP1R12A

www.nature.com/scientificreports

OPEN Transcriptome Analysis Reveals Neuroprotective aspects of Human Reactive Astrocytes induced by Received: 6 February 2017 Accepted: 21 September 2017 Interleukin 1β Published: xx xx xxxx Daniel Boon Loong Teh1, Ankshita Prasad2, Wenxuan Jiang3, Mohd. Zacky Arifn 4, Sanjay Khanna 4, Abha Belorkar5, Limsoon Wong 5, Xiaogang Liu6 & Angelo H. ALL1,7,8

Reactive astrogliosis is a critical process in neuropathological conditions and neurotrauma. Although it has been suggested that it confers neuroprotective efects, the exact genomic mechanism has not been explored. The prevailing dogma of the role of astrogliosis in inhibition of axonal regeneration has been challenged by recent fndings in rodent model’s spinal cord injury, demonstrating its neuroprotection and axonal regeneration properties. We examined whether their neuroprotective and axonal regeneration potentials can be identify in human spinal cord reactive astrocytes in vitro. Here, reactive astrogliosis was induced with IL1β. Within 24 hours of IL1β induction, astrocytes acquired reactive characteristics. Transcriptome analysis of over 40000 transcripts of genes and analysis with PFSnet subnetwork revealed upregulation of chemokines and axonal permissive factors including FGF2, BDNF, and NGF. In addition, most genes regulating axonal inhibitory molecules, including ROBO1 and ROBO2 were downregulated. There was no increase in the gene expression of “Chondroitin Sulfate Proteoglycans” (CSPGs’) clusters. This suggests that reactive astrocytes may not be the main CSPG contributory factor in glial scar. PFSnet analysis also indicated an upregulation of “Axonal Guidance Signaling” pathway. Our result suggests that human spinal cord reactive astrocytes is potentially neuroprotective at an early onset of reactive astrogliosis.

Reactive astrogliosis is a common response of astrocytes to most Central Nervous System (CNS) injury1,2. Te prevailing dogma of the involvement of reactive astrocytes in inhibition of axonal regeneration3, has been challenged by recent fndings of Anderson et al. work, which demonstrated the critical role of reactive astrocytes in aiding axonal regeneration in rodents4. While some of the previous reports have identifed reactive astrocytes as neuroprotective agents1,5,6, their direct neuroprotective efects were not well-documented in the human CNS astrocytes. As astrocytes’ functions difer among regions in the CNS7, we investigated the presence of axonal regenerating and neuroprotective properties of reactive astrocytes4 in human spinal cord derived astrocytes. Elucidating these properties at an early onset of reactive astrogliosis, will enable physicians and scientists to design novel therapeutic strategies exploiting the potential of reactive astrocyte mediated endogenous recovery. It has been reported that among infammatory mediators such as IL-1, TNF, IFN, and TGF, which are known to be active post-CNS injury, cytokine interleukin-1β (IL1β) is specifcally critical for induction of reactive-astrocyte phenotype8. IL1β is a prominent infammatory cytokine and is an early regulator of astrogliosis9. IL1β mRNA is

1Singapore Institute of Neurotechnology (SINAPSE), National University of Singapore, 28 Medical Drive, 5-COR, Singapore, 117456, Singapore. 2Department of Biomedical Engineering, National University of Singapore, E4, 4 Engineering Drive 3, Singapore, 117583, Singapore. 3Department of Orthopaedic Surgery, National University of Singapore, 1E Kent Ridge Road, Singapore, 119228, Singapore. 4Department of Physiology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore. 5Department of Computer Science, National University of Singapore, 13 Computing Drive, Singapore, 117417, Singapore. 6Department of Chemistry, National University of Singapore, 3 Science Drive 3, Singapore, 117543, Singapore. 7Department of Biomedical Engineering and Johns Hopkins School of Medicine, 701C Rutland Avenue 720, Baltimore, MD 21205, USA. 8Department of Neurology, Johns Hopkins School of Medicine, 701C Rutland Avenue 720, Baltimore, MD 21205, USA. Correspondence and requests for materials should be addressed to X.L. (email: [email protected]) or A.H.A. (email: [email protected])

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 1 www.nature.com/scientificreports/

Figure 1. Human spinal cord astrocytes. (A) Control astrocytes stained with GFAP (red) and vimentin (green). (B) IL1β treated human spinal cord reactive astrocytes acquired more extensive processes. (D) Te surface area was signifcantly reduced (p = 0.00863) in reactive astrocytes as compared to control (1200 astrocytes counted for control and treated groups each). (E) Te number of processes to cell ratio was signifcantly increased in reactive astrocytes (p = 9.09 × 10−6, 500 cells were counted each in control and treated groups) afer induced by IL1β. (F) Te processes length was signifcantly increased in reactive astrocyte as compared to control (p = 0.009, 400 processes were counted for experiment and control each). Scale bar 50 μm.

highly up-regulated and persists 24+ hours afer the onset of Spinal Cord Injury (SCI) at and around the epicenter of injury. Tis both initiates and modulates the infammatory responses, leading to reactive astrogliosis10. Although bacterial endotoxin; lipopolysaccharides (LPS) could also be used for inducing reactive astrogliosis11, human astrocytes are unresponsive to LPS stimulation but are highly sensitive to IL1β12. In addition, it has also been reported that clusters of genes expression in reactive-astrocytes were highly dependent on various modali- ties of injury-induction13,14. In our study, we used IL1β to induce reactivity in human spinal cord astrocytes, in vitro within 24 hours. Reactive astrogliosis is known to be invoked at the early phases of injury, with astrocytes acquiring a hypertrophic morphology and increased GFAP expression15–17 (Supp. Figure 1). IL1β treated human spinal cord astrocytes were found to undergo similar morphological transformations within 24 hours of exposure. Using transcriptome analysis, we have identifed that IL6, CXCL5 and C15ORF48 (also known as NMES1 gene) are the most up-regulated genes in human spinal cord reactive astrocytes. Whole genome transcriptome analysis shows changes in genes expression levels of 25 axonal growth permissive and 13 axonal growth inhibitory molecules. Particularly, the axonal growth promotion and neurotrophic factor genes like BDNF and NGF were upregulated. On the other hand, we detected no upregulation of CSPGs clusters of genes, which suggests that reactive astrocytes may not be the major contributors of CSPGs at the early onset (24 hours) of glial scarring. “Axonal Guidance Signaling” and “ECM-Receptor Interaction” pathways in reactive astrocytes, were diferentially upreg- ulates as compared to nascent astrocytes determined by PFSnet subnetwork analysis of diferentially expressed genes (DEGs)18. Collectively, IL1β induced human spinal cord reactive astrocytes may exert various endogenous neuroprotective efects as demonstrated by the upregulation of critical axonal growth genes and downregulation of axonal inhibitory genes. Results Characterization of human spinal cord reactive astrocytes. We tested the homogeneity of the nascent human spinal cord astrocytes by staining with astrocyte markers: Glial Fibrillary Protein (GFAP) and vimentin (Fig. 1A)19–22. Prior to IL1β exposure, the astrocytes were 72 ± 2% positive for GFAP (4075 total cells counted in control group), while all the cells were vimentin+. 24 hours afer exposure to 100 ng/ml of IL1β23,24; the astrocytes acquired bipolar shape and a shrunken morphology with extensive elongated processes (Fig. 1B). Te average surface area of reactive astrocytes was reduced from 2262.6 ± 91 µm2 in control, to 1159.2 ± 52 µm2

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 2 www.nature.com/scientificreports/

No. Gene symbol Fold induction Description 1 IL6 223.01 interleukin 6 2 CXCL5 205.07 chemokine (C-X-C motif) ligand 5 (215101_s_at transcript id) 3 CXCL5 131.80 chemokine (C-X-C motif) ligand 5 (214974_x_at transcript id id) 4 C15orf48 108.20 chromosome 15 open reading frame 48 also known as NMES1 5 CXCL2 98.56 chemokine (C-X-C motif) ligand 2 6 CXCL3 67.81 chemokine (C-X-C motif) ligand 3 7 CCL20 65.40 chemokine (C-C motif) ligand 20 8 CXCL8 53.35 chemokine (C-X-C motif) ligand 8 9 CXCL6 49.07 chemokine (C-X-C motif) ligand 6 10 CXCL1 37.61 chemokine (C-X-C motif) ligand 1 (melanoma growth stimulating activity, alpha) 11 CXCL5 34.83 chemokine (C-X-C motif) ligand 5 12 TNFAIP6 19.95 tumor necrosis factor, alpha-induced protein 6 13 IL1β 18.38 interleukin 1 beta 14 CSF2 17.60 colony stimulating factor 2 (granulocyte-macrophage) 15 MMP3 17.51 matrix metallopeptidase 3 16 BCL2A1 17.06 BCL2-related protein A1 17 CXCL8 14.16 chemokine (C-X-C motif) ligand 8 18 CSF3 13.27 colony stimulating factor 3 19 LOC285628 13.20 MIR146A host gene 20 AREG 12.79 amphiregulin

Table 1. Te 20 most upregulated genes in human spinal cord reactive astrocytes.

in IL1β treated astrocytes (Fig. 1D). Tis change in the surface area was due to the fact that astrocytes acquired a more polarized morphology with extensive processes from the cell bodies. As reported in Fig. 1E, the number of processes to cell ratio for reactive astrocytes (0.25 ± 0) was increased in comparison to control group (0.16 ± 0). Although, a small fraction of control astrocytes displayed extensive processes, their lengths (84.6 ± 5 µm; p = 0.009) were shorter as compared to IL1β treated astrocytes (111.9 ± 5 µm, 400 total processes counted for treated and experiment each) (Fig. 1F).

Genome wide analysis of human spinal cord reactive astrocytes. To investigate reactive astrocytes mediated-endogenous neuroprotection and axonal regeneration potentials, genome wide analysis of reactive astrocytes was carried out. Supp. Figure 2 shows the scatter plot of diferential expression genes between reactive astrocytes and nascent astrocytes. Table 1 shows non-discriminatively the 20 most upregulated genes while, Table 2 shows the top 20 genes downregulated 24 hours afer inducing reactive-astrogliosis. Most prominently, IL6 (223 folds), CXCL5 (205 folds), and C15orf48 (also known as NMES1) (108 folds) were upregulated. EPHA7 was the most downregulated gene, followed by CEMIP and MTUS1 by −20, −18 and −18 folds, respectively. Te complete list of changes in genes expression is provided in Supp. File 1. To elucidate whether GFAP+ or GFAP- astrocytes were the main contributory factor in these transcriptome changes, the fraction of GFAP+ cells were frst determined in control and in reactive astrocytes. Te percentage of GFAP+ cells in reactive astrocytes was increased from 69.0 ± 5% in control to 94.6 ± 0% (p = 0.00152) (Fig. 2A–D,E). Te fraction of IL6+ population in reactive astrocytes were also increased from 28.6 ± 2% to 74.4 ± 1% (Fig. 2A,B,F) (p = 3.4 × 10−6). Likewise, CXCL5 + 6+ population fraction increased by more than 11 folds in reactive astrocytes from control (Fig. 2C,D,F) (p = 0.00571). To determine the distribution of these IL6+ and CXCL5+ 6+ cells among GFAP+/− reactive astrocytes, the fraction of IL6+/GFAP+ and CXCL5+ 6+/GFAP+ reactive astrocytes were tabulated. 94.9 ± 1% GFAP+ reactive astrocytes were also positive with IL6+, while 66.7 ± 4% of GFAP+ reactive astrocytes were also stained with CXCL5+ 6 (Fig. 2G). As a large percentage of GFAP+ astrocytes are double positive for IL6 or CXCL5+ 6 markers, the transcriptome changes can be considered to be a derivation from the GFAP+ reactive astrocytes. Additionally, qRT-PCR was carried out to confrm the increase in IL6 (p = 7.01 × 10−7), CXCL6 (p = 2.34 × 10−6) and NMES1 (p = 6.4 × 10−7). Te fold changes were respectively (IL6) 159.8 ± 37, (CXCL6) 53.0 ± 9 and (NMES1)133.1 ± 15 folds from control (Fig. 2H).

Axonal guidance molecules and neurotrophic factors genes involved in neuroprotection and axonal regeneration. One of the aims of this study was to identify genes regulating axonal guidance molecules and neurotrophic factors expression that promote neuroprotection, neurogenesis and axonal regeneration in human spinal cord reactive astrocytes. Figure 3 shows the fold changes of these specifc genes4. It is noteworthy that 9 out of 25 axonal growth permissive genes were upregulated in reactive astrocytes25–41. Of those, Fibroblast Growth Factor 2 (FGF2) is the most upregulated (3.24 folds), while matrilin2 (MATN2) (−2.37 folds) is the most downregulated axonal permissive genes. On the other hand, Slit Guidance Ligand (SLIT2) (2.54 folds) and Dorsal Inhibitory Axon Guidance Protein (DRAXIN) (2.52 folds) were the most upregulated genes involves in axonal growth inhibitory molecules. 8 out of 13 genes regulating axonal growth inhibitory molecules were down regulated, with Roundabout Guidance Receptor 2 (ROBO2) being the most downregulated genes in reactive astrocytes (−2.5 folds). Additionally, various matrix metallopeptidase and hyaluronan synthases were upregulated

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 3 www.nature.com/scientificreports/

No. Gene symbol Fold induction Description 1 EPHA7 −20.38 EPH receptor A7 2 CEMIP −18.22 cell migration inducing protein, hyaluronan binding 3 MTUS1 −17.57 microtubule associated tumor suppressor 1 4 EPHA7 −17.05 EPH receptor A7 5 COL14A1 −14.93 collagen, type XIV, alpha 1 6 MTUS1 −14.67 microtubule associated tumor suppressor 1 7 ACKR3 −13.86 atypical chemokine receptor 3 8 WNT2B −13.34 wingless-type MMTV integration site family, member 2B 9 PI15 −12.28 peptidase inhibitor 15 10 SLC2A12 −11.74 solute carrier family 2 (facilitated glucose transporter), member 12 11 CXCL12 −11.48 chemokine (C-X-C motif) ligand 12 12 HTR2B −11.25 5-hydroxytryptamine (serotonin) receptor 2B, G protein-coupled 13 TMEM178A −11.03 transmembrane protein 178 A 14 COLEC12 −10.56 collectin sub-family member 12 15 RGCC −9.08 regulator of cell cycle 16 SULF2 −9.00 sulfatase 2 17 SCD −8.90 stearoyl-CoA desaturase (delta-9-desaturase) 18 LRRN1 −8.88 leucine rich repeat neuronal 1 19 SLC2A12 −8.68 solute carrier family 2 (facilitated glucose transporter), member 12 20 SESN3 −8.63 sestrin 3

Table 2. Te 20 most downregulated genes in human spinal cord reactive astrocytes.

(Supp. File 1) as well. Interestingly, CSPGs clusters were unchanged in reactive astrocyte transcriptome as compared to control (Supp. File 1)42–45. Diferential Expressed Genes (DEGs) were then analyzed with KEGG human pathways database to reveal afected biological pathways in human spinal cord reactive astrocytes. Using subnetwork analysis (PFSnet) within each pathway, it is revealed that reactive astrocytes diferentially regulates subnetworks of various pathways, as compared to nascent astrocytes (Fig. 4 and Tables 3–4)46. Consistent with the observable changes in morphology of the in vitro reactive astrogliosis (Figs 1–2), PFSnet analysis revealed that actin cytoskeleton signaling pathway is one of the most altered pathway. As numerous pathways were diferentially regulated in reactive astrocyte, our aim was to focus on main pathways that have critical role in axonal growth and development. We found that reactive astrocytes afect two subnetworks involved in axonal attraction and repulsion; FYN & RASGAP (Ras GTPase-activating protein 1) and SFK (Src family tyrosine kinase). In the “Axonal Guidance Signaling”, the axonal attraction and outgrowth seems to be regulated mainly by FYN and RASGAP (Supp. Figure 3). Another diferential component of axonal attraction appeared to be directly regulated by SFK. In addition, a closely linked pathway to axonal development is the “ECM-receptor interaction” (Supp. Figure 4). Overall, the “ECM-Receptor Interaction” is downregulated in reactive astrocytes. Our result indicates that, fbronectin and laminin were downregulated, while collagen, Trombospondin (THBS), and tenascin were upregulated. Discussion In this study, we report the genome wide transcriptome profle of IL1β induced human spinal cord reactive astrocytes. We have studied a comprehensive subnetwork analysis based transcriptome pattern, which reveals the genes and their associated molecular function along with the important signaling pathways that regulate reactive astrogliosis. Tis provides a molecular roadmap to facilitate intervention and to optimize the benefts of reactive astrogliosis by either augmenting subnetworks of axonal attraction in the “Axonal Guidance Signaling” pathway or inhibiting signaling cascades in the “Axonal Repulsion” subnetwork. CXCL5 and NMES1 are among the most upregulated genes by 205 and 108 folds respectively. Te association of NMES1 with human spinal cord reactive astrocytes is not well-known currently. We speculate that a very likely role of C15orf48 or commonly known as NMES1 may be in the regulation of the cell-cycle process, as its downregulation has been associated with potential tumorigenesis. Similarly, rat model of reactive-astrocyte using lipopolysaccharide (LPS, chemical injury) and middle carotid arterial occlusion (MCAO, ischemic injury) to induce injury did not identify CXCL5 expression in their clusters of diferentially expressed genes. Hence, we deduce that reactive astrocytes are a heterogeneous cell population whose molecular identity is based on the type of injury and origin of host cells13,14,47–49. Te changes in the “Actin Cytoskeleton Signaling” pathway can be attributed to the change of the cytoskeleton into the more difuse and ring-structure actin flaments50, leading to morphology changes in reactive astrocytes14. Te reorganization of the actin cytoskeleton is afected by infammatory cytokines and is closely related to the “Focal Adhesion” and “ECM Receptor Interaction” pathways. A component of MAPK signaling pathway, ERK phosphorylation has been shown to be increased afer SCI51. However, our PFSnet subnetwork analysis suggest that for the “MAPK signaling pathway”, the “JNK and p38 MAPK” sub-networks is activated in reactive astrocytes, while the classical “MAPK kinase sub-pathway” is down regulated. Tis selective subnetwork activation has not been clearly defned previously in reactive astrogliosis. Although the TGF-β signaling pathway is known to

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 4 www.nature.com/scientificreports/

Figure 2. Reactive astrocytes. (A) Control astrocytes stained with IL6 (green), GFAP (red), vimentin (magenta), and DAPI (blue). (B) IL1β reactive astrocytes stained with IL6, GFAP, vimentin and DAPI. (C) Control astrocytes evaluated for CXCL5 + 6 immunocytochemistry, as compared to reactive astrocyte (D). IL1β reactive astrocytes immunocytochemistry staining for CXCL5 + 6 (green). (E) Te fraction of GFAP+ astrocytes population was signifcantly increased as compared to control. (p = 0.00152, 2558 total cells counted in reactive astrocytes and 1962 total cells counted in control) (F) Te overall population fraction positive for either IL6 (p = 3.4 × 10−6, 1299 cells counted in reactive astrocytes, and 965 cells counted in control) or CXCL5 + 6 (p = 0.00571, 1059 cells counted in reactive astrocytes, 1297 cells counted in control) were signifcantly increased in reactive astrocytes as compared to control. (G) Within the GFAP+ cells, 94.9 ± 1% were also co-labelled with IL6, while 66.7 ± 4% were co-labelled with CXCL5 + 6 immunocytochemistry. (H) As a confrmation, a qRT-PCR was carried out for IL6, CXCL6 and NMES1, of which all were upregulated in comparison to control. IL6 (p = 7.01 × 10−7), CXCL6 (p = 2.34 × 10−6) and NMES1 (p = 6.4 × 10−7). Te fold changes were respectively (IL6) 159.8 ± 37, (CXCL6) 53.0 ± 9 and (NMES1)133.1 ± 15 folds from control. Scale bar 20 μm. (n = 4 independent experiments).

Figure 3. Gene fold expression of axonal permissive and inhibitory molecules in reactive astrocytes. 24 genes involved in axonal permissive molecules, were altered in reactive astrocytes from control. On the other hand, 8 out of 13 genes of axonal inhibitory molecules were downregulated. Diferentially Expressed Genes that were listed in modulated biological pathways, were denoted with an arrow showing the biological subnetwork.

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 5 www.nature.com/scientificreports/

Figure 4. Subnetwork analysis in biological pathways in reactive astrocytes. Each major biological pathways are listed and the presence of diferent subnetwork pathways within a major biological pathway are defned by “0, 1, 2, 3, and 4”. Subnetworks are rank based on efect size. Blue graph indicates positively regulated, while red indicates negatively regulated.

be involved in astrocyte reactivity, its exact signaling cascade has not been well-established in reactive astrocytes. Our fndings indicate that one subnetwork of the pathway, involving ERK is upregulated while the other subnetwork involving the RhoA and PP2A is down regulated. Subnetwork analysis ofers the advantage of diferential view within a biological pathway, rather than revealing only an overall general alteration of biological pathway. One intriguing observation was an increase in neurotrophic factors and “Axonal Guidance Signaling” pathway, particularly brain derived neurotrophic factor (BDNF) in human spinal cord reactive astrocytes. Te other neurotrophic factor that was upregulated was nerve growth factor (NGF). Tese two factors are critical in development, survival and regeneration of axons. Our fndings thus suggest that early onset of reactive astrocytes could potentially promote neuronal and axonal development. Reactive astrocytes in the glial scar thus have the potential to secrete multiple axon-growth-permissive molecules, thus improving the microenvironment at and around the epicenter of injury. Tis will facilitate neuronal repair and regeneration. Recent fndings also suggest that BDNF regulates the development of oligodendrocyte precursor cells4,52–56. Our results suggest that the primary axonal attraction aspect of reactive astrocytes, could be attributed to RasGAP and FYN over expression. Additionally, we also report the upregulated of “VEGF signaling” pathway, which has neurotrophic and neuroprotective efects on neuronal cells57. We also identifed that there was an absence of up regulation of CSPGs genes. It has been known that CSPGs are major axonal growth inhibitor in glial scar. Our results present the possibility that reactive astrocytes may not be the main contributor of CPSGs at the epicenter and around the injury. Furthermore, we also reported the upregulation of hyaluronan synthases, which are enzymes for producing hyaluronan. Te overexpression of hyaluronan is known to improve SCI recovery by reducing the lesion, and pro-infammatory cytokines suggesting another neuroprotective properties of reactive astrocytes. Modifcation of extracellular matrix is an important element in formation of glial scar and modulation of axonal growth, regeneration and neuronal development post-CNS injury. MMP3, MMP1, and MMP12 belonging to the cluster of matrix proteins were found to be upregulated in reactive astrocytes. Tese genes have been associated with recruitment and migration of nascent astrocytes to the site of reactive astrocyte. Furthermore, MMP3 and MMP12 overexpression has been associated with remyelination. However, an increase in expression of matrix metalloproteases is known to enhance brain blood barrier (BBB) permeability and immune cells infltration, resulting in infammation post-CNS injury58. Our results indicate the upregulation of collagen and tenascin as well as downregulation of laminin and fbronectin in the ECM-receptor interaction pathways. In particular, collagen is critical in the early phase of tissue repair59 and the role of tenascin, in promoting neural outgrowth is con- troversial. Laminin and fbronectin are well-known for promoting neurite outgrowth. Collectively, we interpret that the ECM changes in reactive astrocytes may potentially induce both neuro-permissive and inhibitory efects to modulate axonal regeneration and growth59–66. In our report, we presented a fetal human spinal cord reactive astrogliosis induced by IL1β. Although the genomic profles of fetal astrocytes and adult astrocytes are comparable, notable diferences in expression has been identifed from previous literature67. Tis include higher expression of pro-infammatory miRNAs expression in adult astrocytes as compared to fetal68, while lower expression of fetal germinal matrix miRNAs in adult astrocytes as compared to fetal67. Te higher expression of matrix associated miRNA could refect the migrating status of the developing astrocytes in fetal CNS. Furthermore, young astrocytes also expressed genes involved

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 6 www.nature.com/scientificreports/

Efect Subnetworks in pathways size Genes PDGFC, SH1, SGK1, GRB2, CDK2, MAP2K3, MAP2K, MAPK1, CDK4, CDK6, MAPK6, Actin Cytoskeleton Signaling_0 2.44 IRAK1, CSNK1A1 Actin Cytoskeleton Signaling_1 5.49 MYH10, PPP1CC, PPP1CA, MYH9, MYLK, MYL6, PPP1R12A, MYL6B Actin Cytoskeleton Signaling_2 9.44 ARHGEF12, RAC1, NCKAP1, RHOA, IQGAP1, PIP4K2C, GNA13 Actin Cytoskeleton Signaling_3 2.33 RRAS, RRAS2, NRAS, PIK3R2, KRAS Antigen processing and presentation_0 16.87 HLA-C, HLA-B, HLA-A, HLA-G, HLA-F, HLA-E EFNB2, ITGB1, PDGFC, CSNK1A1, SGK1, CDK4, SDC2, FYN, NRP1, MAP2K1, GRB2, Axonal Guidance Signaling_0 1.88 VEGFA, PIK3R2, CDK6, MAPK6, MAPK1, RHOD, RASA1, SHC1 Cell cycle_0 2.91 PCNA, CCNB2, CCNB1, GADD45A, CDK4, CDK6, WEE1, YWHAG Colorectal cancer_0 3.47 CCND1, MYC, MAPK1, JUN, MAP2K1 ACTN1, ACTG1, RAC1, VCL, MYLK, ROCK1, MYL6, PPP1R12A, ACTB, RHOA, Focal Adhesion_0 8.49 CDC42 Focal adhesion_1 2.52 PPP1CB, PPP1CC, PPP1CA, ROCK1, PPP1R12A, RHOA ITGB1, ACTN1, PDGFC, ITGB5, SHC1, PRKCA, ITGB8, FYN, PDGFRA, ILK, MET, Focal adhesion_2 1.50 GRB2, VEGFA, PIK3R2, ACTG1, ACTB, PARVA, CAV2, VCL, CAV1 PRKAR2A, SHC1, SGK1, ATF4, CREB3, CDK2, MAP2K3, MAP2K1, GRB2, CDK4, G-protein Coupled Signaling_0 2.77 PRKAR1A, CDK6, MAPK6, MAPK1, IRAK1, CSNK1A1 Insulin signaling pathway_0 3.21 PPP1CB, CALM1, CALM2, PPP1CC, PPP1CA MAPK signaling pathway_0 8.42 HSPA5, HSPA8, HSPA1A, MAPK1, MAPK6, CRK, MYC, DUSP1 MAPK signaling pathway_1 1.14 PRKCA, RRAS2, NRAS, RRAS, RASA1, KRAS MAPK signaling pathway_2 4.12 MKNK2, MAP2K1, MAPK1, ATF4, MYC, DUSP1 Parkinson’s disease_0 8.85 SLC25A6, CYCS, SLC25A5, VDAC3, VDAC2, VDAC1, PPID RRM2, RRM1, ADSL, HPRT1, NT5E, IMPDH2, POLR2H, PAICS, ENTPD4, POLR2E, Purine metabolism_0 3.07 POLR2G, GUK1, POLR2B, POLR2L, ADK, POLR2K, POLR2J, ATIC, NME1, NME2, AK2, POLE3, CANT1, DGUOK RRM2, RRM1, CMPK1, DCTD, NT5E, DTYMK, ENTPD4, POLR2E, POLR2G, POLR2B, Pyrimidine metabolism_0 4.44 POLR2L, DUT, POLR2H, POLR2K, POLR2J, NME1, NME2, AK3, TYMS, ITPA, POLE3, CANT1 CSNK1A1, SGK1, MAPK6, CDK2, MAP2K3, MAP2K1, GRB2, CDK4, CDK6, CRK, sapk-jnk Signaling_0 2.68 MAPK1, IRAK1, SHC1 sapk-jnk Signaling_1 1.66 RRAS, KRAS, RRAS2, NRAS, RAC1 Synaptic Long Term Potentiation_0 2.13 CREB3, RAP1B, RAP1A, PRKAR2A, MAP2K1, MAPK1, PRKAR1A, ATF4 Synaptic Long Term Potentiation_1 5.24 RAP1B, RAP1A, PRKAR2A, MAP2K1, MAPK1, PRKAR1A, ATF4 Synaptic Long Term Potentiation_2 1.35 CALM2, CALM1, PRKCA, RRAS2, NRAS, RRAS, KRAS CSNK1A1, SGK1, GRB2, MAP2K1, MAP2K3, CDK2, MAPK1, CDK4, CDK6, MAPK6, TGF-Beta Signaling_0 2.44 IRAK1 TGF-beta signaling pathway_1 9.59 DCN, TGFBR2, TGFB2, LTBP1, PPP2CA, PPP2CB, THBS1, RHOA Urea cycle and metabolism of amino 2.01 SAT1, AMD1, SMS, ODC1, groups_0 VEGF Signaling_0 15.45 ACTA2, ACTG1, ACTG2, VCL, PRKAR2A, PRKAR1A, ACTB

Table 3. Lists of diferentially expressed genes in each subnetwork of pathways.

in “neuronal diferentiation”69. As our study is based on reactive astrogliosis in fetal astrocytes, it remains to be explored if neuroprotective and axonal regeneration potentials can replicated in adult reactive astrocytes. Also, our analysis indicates that both neuroprotective and inhibitory genes are activated in reactive astrocytes post 24 hours of IL1β exposure, suggesting that a balance of these gene expressions may regulate the functionality of reactive astrocytes. Materials and Methods Cell Culture. Human astrocytes derived from the spinal cord (19 weeks old fetus) was purchased from ScienCell Research Laboratories (Cat. No: 1820)3. The cells were cultured in Astrocytes Basal Medium (ScienCellTM) supplemented with 2% FBS (ScienCellTM), 1% Astrocyte growth supplement (ScienCellTM) and 1% of penicillin-streptomycin (ScienCellTM). Te human astrocytes cells were plated on Matrigel (Corning) coated polystyrene 35mm plates with cell density of 1 × 105 cells. 24 hours afer plating, cells were washed with PBS (HyClone ). Fresh medium containing IL1β (InvitrogenTM) at 100 ng/ml concentration was added and incubated for 24 hours™ in 5% CO2 at 37 °C.

Immunohistochemistry. (i) In vitro: Human spinal cord astrocytes were grown on four wells matrigel coated dishes (Ibidi Inc.) at density of 2 × 104. Tey were fxed with 4% paraformaldehyde (PFA) in PBS for 10 minutes at 4 °C. Te fxative solution was removed and cells were washed three times with PBS for 5 min each at 4 °C. Cells were permeabilized and blocked with blocking bufer for 30 minutes at 4 °C. Primary antibodies (1:200; GFAP) [Millipore], (1:50; IL-6) [sc-1265 Santa Cruz], (1:50; Cxcl5 + 6) [ab198505, Abcam] and (1:250;Vimentin) [RD system] was diluted in blocking bufer and incubated overnight at 4 °C. Samples were then

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 7 www.nature.com/scientificreports/

washed three times with washing bufer for 10 minutes each at 4 °C. Secondary antibodies were added (Alexa 488, Alexa 564 and Streptavidin 643 conjugated antibodies, 1:200) in blocking bufer and cells were incubated at room for 4 hours. Cells were then washed three times with washing bufer for 5 minutes each at room temperature. Nuclei were stained with DAPI for 10 min at room temperature. DAPI solution was aspirated and washed once washing bufer for 5 minutes at room temperature was performed. Mounting medium DAKO was added and the samples were incubated at 4 °C overnight.

RNA isolation and microarray. Cells were collected after exposure to IL1β for 24 hours, Accutase (StemPro ) was added to detach the cells. RNA was extracted using RNeasy Mini Kit (Qiagen, CA, USA) per manufacturer® specifcations.

Microarray. (i) All RNA samples were quality checked prior to microarray analysis in accordance to Affymetrix recommended protocols. Affymetrix 3′ IVT PLUS reagent kit was used. All RNA samples were assessed with spectrophotometric measurements (BioSPEC-Mini, Shimadzu) and quality RNA Integrity Number (RIN) with Agilent Bioanalyzer, (ii) for target preparation, 100 ng of total RNA was reverse transcribed to generate cDNA, and later used as template to generate biotin-labeled amplifed RNA (aRNA). aRNA was then fragmented and hybridized to Afymetrix Human U133 Plus 2.0 Arrays, for 16 hours (45 °C; 60 rpm rotation). Afymetrix Human U133 Plus 2.0 Arrays (HG-U133 Plus 2.0) comprised of 1,300,000 unique oligonucleotides, encompass- ing 47, 000 transcripts and variants, representing approximately 39, 000 of well characterized human genes, pro- viding a complete coverage to the human genes. All samples were processed with similar reagent kit, were washed, stained and scanned with Afymetrix 300 7 G scanner. Scanned images were assessed for hybridization efciency. (iii) Prior to microarray analysis, quality control checks were also carried out on the array. Te 3′/5′ ratio of housekeeping genes, presences of spike controls, background value, raw Q noise, scaling factor, and percent of genes presence were evaluated. Signal intensity ratio of 3′/5′ probe sets is important in establishing good cDNA synthesis, integrity of starting RNA and hybridization properties. In order to obtain reliable and accurate data comparison, the background intensities were measured and were made sure to be consistent and closed to each other. Additionally, the raw Q noise was also measured by performing pixel to pixel variations in background intensities. Tis is to make sure that the variations in the digitized signal observed by the scanners as it samples the probe array’s surface is consistent and close to each other. Te measure of brightness of the array which may vary between arrays to array was also normalized to a standard level. Tis normalization will be important in comparing array to array data. To assess the hybridization, washing and staining steps quality, bacteria spike controls were performed. Tese are probe sets hybridized by pre-labeled bacterial spike controls (BioC, BioC, BioD, and Cre in staggered concentration). Finally, Poly A control was also carried out to identify problems with target preparation. Tese are a set of poly-adenylated RNA spikes of Lys, Phe, Tr, and Dap, prepared in staggered concentration. Tese are prepared together with the samples throughout cDNA synthesis onwards. All RNA samples have RIN 10 and good quality absorbance (OD) ratio for 260/280 m, (>1.8) and 260/230 (>2.0) (Supp. Table 1, Supp Fig. 5), suggesting high purity of RNA, with minimal degradation. Furthermore, the 3′/5′ ratio of housekeeping gene GAPDH, approximated 1 onwards (Supp. Table 2). Te array was also quality checked with consistent background intensity ranging from 30.781324 to 40.71225 (Supp. Table 2). Te normalized scaling factors was done to standardize the measure of brightness of the array, and the result of the array falls within the acceptable range of below 3-fold (Supp. Table 2). In the bacteria spike controls, all samples also showed correct signal intensities of the control including BioB, suggesting good hybridization (Supp. Fig. 6). To rule out any problems with the target preparation, PolyA controls were also carried out, and all samples have shown consistent and staggered concentration of the poly A controls (Supp. Fig. 6).

Analysis of array intensities. For the changes in differential genes expression level, Affymetrix Transcriptome Console was used to analyze the level of expression for each individual gene. One-way between subject ANOVA, was used to assess the statistical signifcance (p < 0.05).

Quantitative Reverse-Transcriptase Polymerase Chain Reaction. qRT-PCR was performed on ViiA 7 Real-Time PCR System (Applied Biosystems ) using SYBR Green PCR Master Mix reagent (Applied™ Biosystems ). Specific PCR products for™ IL-6 (Fwd-ATAGCCCAGAGCATCCCTCC, Rev-GGGTCAGGGGTGGTTATTGC),™ GFAP (Fwd-CAGATTCGAGGGGGCAAAAGC, Rev-AGGCTCACTCCCTGTCAAGC), and Cxcl 6 (Fwd-TGCGTTGCACTTGTTTACGC, Rev-CTTCCCGTTCTTCAGGGAGG) NMES1 (Fwd-GCCCACCAGGCGATCAATAC, Rev-ACACAGCGAAAGATGAGGCT) were detected with the fuorescent double-stranded DNA binding dye, SYBR Green. qRT-PCR amplifcation was performed in triplicates for each sample and the results were replicated in four independent experiments. Gel electrophoresis and melting curve analyses were performed to validate PCR product sizes. Te expression level of each gene was normalized against β-actin using the comparative CT method70.

Subnetwork analysis. Candidate subnetworks were generated by inducing connected components on known biological pathways with highly expressed genes in each phenotype. Two scores are computed for each subnetwork; these scores denote the level of expression of the subnetwork in majority of the samples in each phenotype. Finally, the diference of the two scores is tested for statistical signifcance. Te theoretical t-distribution is used as the null distribution for estimating the statistical signifcance of subnetworks scored in PFSNet. An additional criteria set was that any subnetwork tested for statistical signifcance, has to be highly expressed in all the samples in the corresponding group (control/IL1β). Changed in subnetworks were analyzed with pathway information from the PathwayAPI database34 which contains the aggregation of human pathways from KEGG71 and Ingenuity. Expression data is preprocessed using the espresso function in R afy package336. Pathways maps

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 8 www.nature.com/scientificreports/

Subnetworks in pathways Efect size Genes ACTN1, TJP1, ACP1, ACTB, RAC1, IQGAP1, VCL, FYN, CSNK2B, CTNNB1, MET, Adherens junction_0 2.12 CDC42, PTPRF, RHOA, PTPRM, CTNNA1, CSNK2A1, ACTG1 Alanine and aspartate metabolism_0 1.20 DARS, GOT2, ASS1, ASNS, NARS Biosynthesis of steroids_0 1.66 FDPS, IDI1, LSS, CYP51 A1, FDFT1, SQLE Butanoate metabolism_0 1.87 ECHS1, ACAT2, HADHA, HSD17B4, ACAT1 Calcium signaling pathway_0 1.34 CAMK2D, PPP3CB, CALM2, MYLK, CALM1 Calcium Signaling_0 4.89 RAP2B, CALM2, CALM1, RAP1B, RAP1A, PRKAR2A, MAPK1, PRKAR1A, ATF4 IDH3B, IDH3G, DLD, IDH2, IDH1, ACO1, SUCLG1, MDH2, MDH1, DLST, CS, ACLY, Citrate cycle (TCA cycle)_0 3.55 PDHB, SDHA, PDHA1, SDHD, FH, SDHC LAMC1, COL11A1, FN1, CD44, SDC2, SDC4, THBS1, COL5A2, COL6A1, COL5A1, ECM-receptor interaction_0 1.57 COL1A2, COL4A2, COL4A1, COL1A1, LAMB1, COL3A1, COL6A3, TNC, CD47 Fatty acid metabolism_0 1.59 ACADVL, ECHS1, HADHB, ACSL3, HADHA, ACAT1, ACAT2, HSD17B4 Fructose and mannose metabolism_0 11.75 PHPT1, HK1, TPI1, PFKP, PFKM, ALDOA Galactose metabolism_0 1.76 AKR1B1, HK1, PGM1, UGP2, GLB1 GJA1, TJP1, TUBA1C, TUBA1B, TUBA1A, TUBB, TUBB3, TUBB6, MAP2K1, MAPK1, Gap junction_0 2.14 PRKCA, TUBB2A, TUBB2B Glutamate metabolism_0 3.50 GLS, GLUD1, GLUL, GOT2, EPRS, QARS Glutathione metabolism_0 2.48 GSTO1, GSTA4, GSTM3, TXNDC12, GPX1, MGST3, GSTP1, MGST1, GPX4 Glycine, serine and threonine 3.00 PSPH, SHMT2, PSAT1, PHGDH, SARS, GARS metabolism_0 Glycogen Metabolism_0 1.53 CALM2, CALM1, PPP2CA, PPP2CB, PGM1, PYGB Glycolysis / Gluconeogenesis_0 3.78 PGM1, PFKM, GPI, HK1, GAPDH, PFKP, ENO1, TPI1, PGAM1, ALDOA, PGK1 Glycolysis / Gluconeogenesis_1 13.34 PDHB, PDHA1, LDHA, LDHB, DLD IL-2 Signaling_0 2.75 PTPN11, GRB2, PIK3R2, SHC1, JAK1 Integrin Signaling_0 1.82 ACTA2, ACTG1, ACTG2, RHOQ, RND3, RHOC, RHOA, ACTB, RHOD Jak-STAT signaling pathway_0 3.88 PTPN11, IL6ST, GRB2, PIK3R2, IFNGR2, JAK1, OSMR Lysine degradation_0 1.87 ACAT1, ACAT2, HADHA, HSD17B4, ECHS1 Metabolism of xenobiotics by 3.02 GSTO1, GSTA4, GSTM3, MGST3, GSTP1, MGST1 cytochrome P450_0 Pentose phosphate pathway_0 3.78 GPI, TALDO1, PGM1, PFKP, TKT, ALDOA, PFKM Phosphatidylinositol signaling 3.24 PTEN, CDIPT, PIK3R2, PIP4K2C, SYNJ2 system_0 Proteasome Degradation_0 3.50 PSME2, PSME1, PSMB7, PSMB6, PSMB5, PSMB4, PSMB3, PSMB2, PSMB1 Pyruvate metabolism_0 10.63 PDHA1, DLD, ME1, MDH2, MDH1, PDHB, LDHA, LDHB Reductive carboxylate cycle (CO2 2.07 IDH2, IDH1, FH, MDH2, MDH1, ACO1, ACLY fxation)_0 Signaling of Hepatocyte Growth 14.67 MAP2K1, MAPK1, JUN, RAP1B, RAP1A Factor Receptor_0 Starch and sucrose metabolism_0 1.88 GPI, UXS1, UGDH, HK1, PGM1, PYGB, GBE1, UGP2 PPP1R7, PPP1R3C, PPP1CC, PPP1CA, PRKCA, PPP2CB, PPP2CA, MAP2K1, MAPK1, Synaptic Long Term Depression_0 1.62 PPP1R12A, PPP1R14B Synaptic Long Term Depression_1 2.10 YWHAZ, GNAS, PRDX6, GNA13, GRN, GNAI3 MYH10, TJP1, ACTG1, PRKCA, MYH9, CTNNA1, CTNNB1, CTTN, CLDN11, ACTB, Tight junction_0 2.48 RHOA Translation Factors_0 4.53 EIF1, EIF1AX, EIF5B, EIF4H, EIF4B, EIF4A1, EIF4A2 Tryptophan metabolism_0 1.87 ACAT1, ACAT2, HADHA, HSD17B4, ECHS1 Valine, leucine and isoleucine 1.73 ALDH7A1, ECHS1, ACAT2, HADHA, HSD17B4, ACAT1, HADHB, ALDH9A1 degradation_0 wnt Signaling_0 0.98 TCF4, TCF3, GJA1, CCND1, CD44, MYC

Table 4. Lists of genes for pathways downregulated in reactive astrocytes.

were generated based on KEGG pathways database tool3. Each subnetwork was ranked by their efect size. Efect size is a quantitative measure of diference between two groups. In our study, the efect size for each subnetwork was computed as the standardized mean of paired diference “(mean of paired diferences)/(standard deviation of paired diferences)” between PFSNet scores, corresponding to the control and reactive astrocyte groups72,73. As the sample size is small, multiple-testing correction was not performed since the p-value of the same gene fuctu- ates in a small range, with a large portion in the insignifcant part even though the gene is diferentially expressed by construction. Multiple-testing correction approaches would simply shif the null-hypothesis rejection threshold lef-wards and thus would be insensitive against the wide fuctuation range in the p-value of this gene. We have also discuss similar approach previously62,63.

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 9 www.nature.com/scientificreports/

Imaging. Te samples were imaged using ZEISS LSM 810 confocal microscopy instrument. Image J (Fiji) was used for image analysis and quantifcation. Tile scan and stitching were performed with ZEN blue sofware to image the entire spinal cord slice. Quantifcation for spinal cord GFAP intensity was done over 5 random Region of Interest (ROIs), covering the grey and white matter. Te ROI is a 200 × 200 μm2 square visualized with (Plan-APOCHROMAT Zeiss NA = 0.45) 10X objective lens. 3 slices with 400 um between each of them were used for each rat. For in vitro cell length and processes quantifcation, 20X Plan-APOCHROMAT (NA = 0.8) objective lens was used. Cells that displayed processes were selected for quantifcation since not all the cells displayed a morphology that included processes. Te main primary process of the cell was defned as any extension protruding from the cell body. Te length of the primary process was defned as the distance from the center of the nuclei to the tip of the extended process. Only the main primary processes were used for quantifcation since very few cells displayed branching processes afer exposure to IL-1B. For morphological quantifcation, the gray- scale images were converted to binary type using ImageJ and their threshold was adjusted to clearly demarcate the cell boundary. Te wand tool was used to outline the cellular perimeter in order to measure the cell surface area. Measuring was done using the line tool in Image J. Imaging settings were fxed for all groups within each experiments.

Statistical analysis. All results were expressed as mean ± SEM, unless stated. Experiments were repeated independently for four times (n = 4) unless stated. Statistical signifcance was evaluated with unpaired student t-test for signifcance (*p < 0.05; **p < 0.01). References 1. Sofroniew, M. V. Molecular dissection of reactive astrogliosis and glial scar formation. Trends Neurosci 32, 638–647, https://doi. org/10.1016/j.tins.2009.08.002 (2009). 2. Paniagua-Torija, B., Arevalo-Martin, A., Molina-Holgado, E., Molina-Holgado, F. & Garcia-Ovejero, D. Spinal cord injury induces a long-lasting upregulation of interleukin-1beta in astrocytes around the central canal. Neuroscience 284, 283–289, https://doi. org/10.1016/j.neuroscience.2014.10.013 (2015). 3. Linker, R. A. et al. Fumaric acid esters exert neuroprotective efects in neuroinfammation via activation of the Nrf2 antioxidant pathway. Brain. 134, 678–92 (2011). 4. Anderson, M. A. et al. Astrocyte scar formation aids central nervous system axon regeneration. Nature 532, 195-200, doi:https://doi. org/10.1038/nature17623, http://www.nature.com/nature/journal/v532/n7598/abs/nature17623.html#supplementary-information (2016). 5. Faulkner, J. R. et al. Reactive astrocytes protect tissue and preserve function afer spinal cord injury. J Neurosci 24, 2143–2155, https://doi.org/10.1523/jneurosci.3547-03.2004 (2004). 6. Sofroniew, M. V. & Vinters, H. V. Astrocytes: biology and pathology. Acta Neuropathol 119, 7–35, https://doi.org/10.1007/s00401- 009-0619-8 (2010). 7. Oberheim, N. A., Goldman, S. A. & Nedergaard, M. Heterogeneity of astrocytic form and function. Methods Mol Biol 814, 23–45, https://doi.org/10.1007/978-1-61779-452-0_3 (2012). 8. Wang, Y. et al. Interleukin-1beta induces blood-brain barrier disruption by downregulating Sonic hedgehog in astrocytes. PLoS One 9, e110024, https://doi.org/10.1371/journal.pone.0110024 (2014). 9. Herx, L. M. & Yong, V. W. Interleukin-1β is Required for the Early Evolution of Reactive Astrogliosis Following CNS Lesion. Journal of Neuropathology & Experimental Neurology 60, 961–971, https://doi.org/10.1093/jnen/60.10.961 (2001). 10. Sato, A. et al. Interleukin-1 participates in the classical and alternative activation of microglia/macrophages afer spinal cord injury. J Neuroinfammation 9, 65, https://doi.org/10.1186/1742-2094-9-65 (2012). 11. Santos-Galindo, M., Acaz-Fonseca, E., Bellini, M. J. & Garcia-Segura, L. M. Sex diferences in the infammatory response of primary astrocytes to lipopolysaccharide. Biol Sex Difer 2, 7, https://doi.org/10.1186/2042-6410-2-7 (2011). 12. Tarassishin, L., Suh, H. S. & Lee, S. C. LPS and IL-1 diferentially activate mouse and human astrocytes: role of CD14. Glia 62, 999–1013, https://doi.org/10.1002/glia.22657 (2014). 13. John, G. R. et al. Multiple sclerosis: re-expression of a developmental pathway that restricts oligodendrocyte maturation. Nat Med 8, 1115–1121, https://doi.org/10.1038/nm781 (2002). 14. Zamanian, J. L. et al. Genomic analysis of reactive astrogliosis. J Neurosci 32, 6391–6410, https://doi.org/10.1523/ jneurosci.6221-11.2012 (2012). 15. Wu, D., Shibuya, S., Miyamoto, O., Itano, T. & Yamamoto, T. Increase of NG2-positive cells associated with radial glia following traumatic spinal cord injury in adult rats. J Neurocytol 34, 459–469, https://doi.org/10.1007/s11068-006-8998-4 (2005). 16. Gao, D. et al. Te molecular cloning of glial fbrillary acidic protein in Gekko japonicus and its expression changes afer spinal cord transection. Cellular & Molecular Biology Letters 15, 582–599, https://doi.org/10.2478/s11658-010-0029-x (2010). 17. M’Dahoma, S. et al. Spinal Cord Transection-Induced Allodynia in Rats – Behavioral, Physiopathological and Pharmacological Characterization. PLoS ONE 9, e102027, https://doi.org/10.1371/journal.pone.0102027 (2014). 18. Lim, K. & Wong, L. Finding consistent disease subnetworks using PFSNet. Bioinformatics 30, 189–196, https://doi.org/10.1093/ bioinformatics/btt625 (2014). 19. Schnitzer, J., Franke, W. W. & Schachner, M. Immunocytochemical demonstration of vimentin in astrocytes and ependymal cells of developing and adult mouse nervous system. Te Journal of cell biology 90, 435–447 (1981). 20. Chandrasekaran, A., Avci, H. X., Leist, M., Kobolak, J. & Dinnyes, A. Astrocyte Diferentiation of Human Pluripotent Stem Cells: New Tools for Neurological Disorder Research. Front Cell Neurosci 10, 215, https://doi.org/10.3389/fncel.2016.00215 (2016). 21. Roybon, L. et al. Human stem cell-derived spinal cord astrocytes with defned mature or reactive phenotypes. Cell Rep 4, 1035–1048, https://doi.org/10.1016/j.celrep.2013.06.021 (2013). 22. Beaudet, M.-J. et al. High yield extraction of pure spinal motor neurons, astrocytes and microglia from single embryo and adult mouse spinal cord. 5, 16763, https://doi.org/10.1038/srep16763, https://www.nature.com/articles/srep16763#supplementary- information (2015). 23. Molina-Holgado, F., Toulmond, S. & Rothwell, N. J. Involvement of interleukin-1 in glial responses to lipopolysaccharide: endogenous versus exogenous interleukin-1 actions. J Neuroimmunol 111, 1–9 (2000). 24. von Boyen, G. B. et al. Proinfammatory cytokines increase glial fbrillary acidic protein expression in enteric glia. Gut 53, 222–228 (2004). 25. Walsh, F. S. & Doherty, P. Neural cell adhesion molecules of the immunoglobulin superfamily: role in axon growth and guidance. Annu Rev Cell Dev Biol 13, 425–456, https://doi.org/10.1146/annurev.cellbio.13.1.425 (1997). 26. Dickson, B. J. Molecular mechanisms of axon guidance. Science 298, 1959–1964, https://doi.org/10.1126/science.1072165 (2002). 27. Szebenyi, G. et al. Fibroblast growth factor-2 promotes axon branching of cortical neurons by infuencing morphology and behavior of the primary growth cone. J Neurosci 21, 3932–3941 (2001).

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 10 www.nature.com/scientificreports/

28. Rajagopalan, S. et al. Neogenin mediates the action of repulsive guidance molecule. Nat Cell Biol 6, 756–762, https://doi.org/10.1038/ ncb1156 (2004). 29. Tom, V. J., Doller, C. M., Malouf, A. T. & Silver, J. Astrocyte-associated fbronectin is critical for axonal regeneration in adult white matter. J Neurosci 24, 9282–9290, https://doi.org/10.1523/jneurosci.2120-04.2004 (2004). 30. Harel, N. Y. & Strittmatter, S. M. Can regenerating axons recapitulate developmental guidance during recovery from spinal cord injury? Nat Rev Neurosci 7, 603–616, https://doi.org/10.1038/nrn1957 (2006). 31. Nakanishi, K. et al. Identifcation of neurite outgrowth-promoting domains of neuroglycan C, a brain-specifc chondroitin sulfate proteoglycan, and involvement of phosphatidylinositol 3-kinase and protein kinase C signaling pathways in neuritogenesis. J Biol Chem 281, 24970–24978, https://doi.org/10.1074/jbc.M601498200 (2006). 32. Ozdinler, P. H. & Macklis, J. D. IGF-I specifically enhances axon outgrowth of corticospinal motor neurons. Nat Neurosci 9, 1371–1381, https://doi.org/10.1038/nn1789 (2006). 33. Yang, Z. et al. NG2 glial cells provide a favorable substrate for growing axons. J Neurosci 26, 3829–3839, https://doi.org/10.1523/ jneurosci.4247-05.2006 (2006). 34. Lu, P., Jones, L. L. & Tuszynski, M. H. Axon regeneration through scars and into sites of chronic spinal cord injury. Exp Neurol 203, 8–21, https://doi.org/10.1016/j.expneurol.2006.07.030 (2007). 35. Minor, K. et al. Decorin promotes robust axon growth on inhibitory CSPGs and myelin via a direct efect on neurons. Neurobiol Dis 32, 88–95, https://doi.org/10.1016/j.nbd.2008.06.009 (2008). 36. Plantman, S. et al. Integrin-laminin interactions controlling neurite outgrowth from adult DRG neurons in vitro. Mol Cell Neurosci 39, 50–62, https://doi.org/10.1016/j.mcn.2008.05.015 (2008). 37. White, R. E., Yin, F. Q. & Jakeman, L. B. TGF-alpha increases astrocyte invasion and promotes axonal growth into the lesion following spinal cord injury in mice. Exp Neurol 214, 10–24, https://doi.org/10.1016/j.expneurol.2008.06.012 (2008). 38. Alto, L. T. et al. Chemotropic guidance facilitates axonal regeneration and synapse formation afer spinal cord injury. Nat Neurosci 12, 1106–1113, https://doi.org/10.1038/nn.2365 (2009). 39. Malin, D. et al. Te extracellular-matrix protein matrilin 2 participates in peripheral nerve regeneration. J Cell Sci 122, 995–1004, https://doi.org/10.1242/jcs.040378 (2009). 40. Ahmed, G. et al. Draxin inhibits axonal outgrowth through the netrin receptor DCC. J Neurosci 31, 14018–14023, https://doi. org/10.1523/jneurosci.0943-11.2011 (2011). 41. Selvaraj, P. et al. Neurotrophic factor-alpha1 modulates NGF-induced neurite outgrowth through interaction with Wnt-3a and Wnt- 5a in PC12 cells and cortical neurons. Mol Cell Neurosci 68, 222–233, https://doi.org/10.1016/j.mcn.2015.08.005 (2015). 42. Klapka, N. & Muller, H. W. Collagen matrix in spinal cord injury. J Neurotrauma 23, 422–435, https://doi.org/10.1089/ neu.2006.23.422 (2006). 43. Hill, J. J., Jin, K., Mao, X. O., Xie, L. & Greenberg, D. A. Intracerebral chondroitinase ABC and heparan sulfate proteoglycan glypican improve outcome from chronic stroke in rats. Proc Natl Acad Sci USA 109, 9155–9160, https://doi.org/10.1073/pnas.1205697109 (2012). 44. Edwards, T. J. & Hammarlund, M. Syndecan promotes axon regeneration by stabilizing growth cone migration. Cell Rep 8, 272–283, https://doi.org/10.1016/j.celrep.2014.06.008 (2014). 45. Farhy Tselnicker, I., Boisvert, M. M. & Allen, N. J. Te role of neuronal versus astrocyte-derived heparan sulfate proteoglycans in brain development and injury. Biochem Soc Trans 42, 1263–1269, https://doi.org/10.1042/bst20140166 (2014). 46. Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28, 27–30 (2000). 47. Dawes, J. M. et al. CXCL5 Mediates UVB Irradiation–Induced Pain. Science translational medicine 3, 90ra60–90ra60, https://doi. org/10.1126/scitranslmed.3002193 (2011). 48. Spisak, S. et al. Genome-wide screening of genes regulated by DNA methylation in colon cancer development. PLoS One 7, e46215, https://doi.org/10.1371/journal.pone.0046215 (2012). 49. Liu, X. et al. Stat3 inhibition attenuates mechanical allodynia through transcriptional regulation of chemokine expression in spinal astrocytes. PLoS One 8, e75804, https://doi.org/10.1371/journal.pone.0075804 (2013). 50. Hansson, E. Actin flament reorganization in astrocyte networks is a key functional step in neuroinfammation resulting in persistent pain: novel fndings on network restoration. Neurochem Res 40, 372–379, https://doi.org/10.1007/s11064-014-1363-6 (2015). 51. Ito, M. et al. Type I interferon inhibits astrocytic gliosis and promotes functional recovery afer spinal cord injury by deactivation of the MEK/ERK pathway. J Neurotrauma 26, 41–53, https://doi.org/10.1089/neu.2008.0646 (2009). 52. Harlow, E. G. et al. Critical period plasticity is disrupted in the barrel cortex of FMR1 knockout mice. Neuron 65, 385–398, https:// doi.org/10.1016/j.neuron.2010.01.024 (2010). 53. Chen, A., Xiong, L. J., Tong, Y. & Mao, M. Te neuroprotective roles of BDNF in hypoxic ischemic brain injury. Biomed Rep 1, 167–176, https://doi.org/10.3892/br.2012.48 (2013). 54. Baydyuk, M. & Xu, B. BDNF signaling and survival of striatal neurons. Frontiers in Cellular Neuroscience 8, 254, https://doi. org/10.3389/fncel.2014.00254 (2014). 55. Min, S. K. et al. Assessment of C-phycocyanin efect on astrocytes-mediated neuroprotection against oxidative brain injury using 2D and 3D astrocyte tissue model. Scientifc Reports 5, 14418, https://doi.org/10.1038/srep14418 http://www.nature.com/articles/ srep14418#supplementary-information (2015). 56. Tsiperson, V. et al. Brain-derived neurotrophic factor defciency restricts proliferation of oligodendrocyte progenitors following cuprizone-induced demyelination. ASN Neuro 7, https://doi.org/10.1177/1759091414566878 (2015). 57. Zachary, I. Neuroprotective role of vascular endothelial growth factor: signalling mechanisms, biological function, and therapeutic potential. Neuro-Signals 14, 207–221, https://doi.org/10.1159/000088637 (2005). 58. Mirshafey, A., Asghari, B., Ghalamfarsa, G., Jadidi-Niaragh, F. & Azizi, G. Te Signifcance of Matrix Metalloproteinases in the Immunopathogenesis and Treatment of Multiple Sclerosis. Sultan Qaboos University Medical Journal 14, e13–e25 (2014). 59. Teller, P. & White, T. K. Te physiology of wound healing: injury through maturation. Surg Clin North Am 89, 599–610, https://doi. org/10.1016/j.suc.2009.03.006 (2009). 60. Jankovski, A. & Sotelo, C. Subventricular zone-olfactory bulb migratory pathway in the adult mouse: cellular composition and specificity as determined by heterochronic and heterotopic transplantation. J Comp Neurol 371, 376–396, doi: https://doi. org/10.1002/(SICI)1096-9861(19960729)371:3<376::AID-CNE3>3.0.CO;2-# (1996). 61. Zhang, J. et al. Enhancement of rat sciatic nerve regeneration by fbronectin and laminin through a silicone chamber. J Reconstr Microsurg 19, 467–472, https://doi.org/10.1055/s-2003-44635 (2003). 62. Allan, S. M., Tyrrell, P. J. & Rothwell, N. J. Interleukin-1 and neuronal injury. Nat Rev Immunol 5, 629–640, https://doi.org/10.1038/ nri1664 (2005). 63. Skuljec, J. et al. Matrix metalloproteinases and their tissue inhibitors in cuprizone-induced demyelination and remyelination of brain white and gray matter. J Neuropathol Exp Neurol 70, 758–769, https://doi.org/10.1097/NEN.0b013e3182294fad (2011). 64. Austin, J. W., Gilchrist, C. & Fehlings, M. G. High molecular weight hyaluronan reduces lipopolysaccharide mediated microglial activation. J Neurochem 122, 344–355, https://doi.org/10.1111/j.1471-4159.2012.07789.x (2012). 65. Gaudet, A. D. & Popovich, P. G. Extracellular matrix regulation of infammation in the healthy and injured spinal cord. Exp Neurol 258, 24–34, https://doi.org/10.1016/j.expneurol.2013.11.020 (2014). 66. Wang, S. M. et al. Astrocytic CCAAT/Enhancer-Binding Protein Delta Contributes to Glial Scar Formation and Impairs Functional Recovery Afer Spinal Cord Injury. Mol Neurobiol 53, 5912–5927, https://doi.org/10.1007/s12035-015-9486-6 (2016).

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 11 www.nature.com/scientificreports/

67. Rao, V. T. et al. MicroRNA Expression Patterns in Human Astrocytes in Relation to Anatomical Location and Age. J Neuropathol Exp Neurol 75, 156–166, https://doi.org/10.1093/jnen/nlv016 (2016). 68. Lafortune, L., Nalbantoglu, J. & Antel, J. P. Expression of tumor necrosis factor alpha (TNF alpha) and interleukin 6 (IL-6) mRNA in adult human astrocytes: comparison with adult microglia and fetal astrocytes. J Neuropathol Exp Neurol 55, 515–521 (1996). 69. Orre, M. et al. Acute isolation and transcriptome characterization of cortical astrocytes and microglia from young and aged mice. Neurobiology of aging 35, 1–14, https://doi.org/10.1016/j.neurobiolaging.2013.07.008 (2014). 70. Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402–408, https://doi.org/10.1006/meth.2001.1262 (2001). 71. Gautier, L., Cope, L., Bolstad, B. M. & Irizarry, R. A. Afy–analysis of Afymetrix GeneChip data at the probe level. Bioinformatics 20, 307–315, https://doi.org/10.1093/bioinformatics/btg405 (2004). 72. Goh, W. W. B. & Wong, L. Advancing Clinical Proteomics via Analysis Based on Biological Complexes: A Tale of Five Paradigms. Journal of Proteome Research 15, 3167–3179, https://doi.org/10.1021/acs.jproteome.6b00402 (2016). 73. Wang, W., Sue, A. C. H. & Goh, W. W. B. Feature selection in clinical proteomics: with great power comes great reproducibility. Drug Discovery Today, https://doi.org/org/10.1016/j.drudis.2016.12.006. Acknowledgements We would like to thank Ms. Low Kay En and Ms Tan Soon Huat from Electron Microscopy Unit, Yong Loo Lin School of Medicine, and Ms. Tan Shu Ying from the Confocal Microscopy Unity of National University of Singapore, for their technical assistance.This work is supported by the R-175-000-137-112 (Singapore’s Ministry of Education Tier 1), R-175-000-121-733 (Singapore Institute for Neurotechnology Seed Fund), (2013-MSCRFII-0109-00, PI: A. ALL) State of Maryland, USA. Author Contributions Experimental design: A.H.A., X.L., D.B.T., K.S. Experiments carried out by: D.B.T., A.P., A.M.Z., J.W., Data analysis: D.B.T., A.P., A.H.A., L.W., B.A., J.W., Manuscript writing: D.B.T., A.P., B.A., L.W., K.S., X.L., A.H.A., Manuscript revision: All authors. Additional Information Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-13174-w. Competing Interests: Te authors declare that they have no competing interests. Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre- ative Commons license, and indicate if changes were made. Te images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Scientific REPOrTS | 7:13988 | DOI:10.1038/s41598-017-13174-w 12